Product-Market Fit: Measure It Like You Mean It

The 40% test, retention curves, and growth accounting. PMF is a spectrum, not a switch. Here's how to know where you stand.

By Prateek Jain
9 min readAdvanced

Prerequisites

  • • Understanding of core product metrics
  • • Basic knowledge of customer research methods

What PMF Actually Is

Marc Andreessen described PMF as follows1:

"Product-market fit means being in a good market with a product that can satisfy that market."

PMF is not binary. You can see early evidence in one segment and little evidence in another, and market conditions can change over time. Use multiple quantitative and qualitative signals rather than treating any one observation as proof.

CB Insights lists no market need among reported reasons for startup failure, but that category does not imply that all other companies have achieved PMF.2

The PMF Spectrum

These four stages offer a way to organize the evidence:

StageSignalWhat to do
No PMFLittle evidence of lasting customer valueTest new hypotheses.
Early PMFSome segments show sustained demandInvestigate those segments and their needs.
Strong PMFSustained demand supported by retention and feedbackStrengthen infrastructure and assess expansion.
Scaling PMFDemand exceeds current delivery capacityExpand capacity while monitoring customer value.

The transition from early to stronger PMF is a judgment based on the evidence, customer segment, and business model.

The Five Quantitative Metrics

Use these five metrics alongside customer research; they answer different questions.

1. The Sean Ellis 40% Test

Ask active users one question3:

"How would you feel if you could no longer use [product]?"

  • Very disappointed
  • Somewhat disappointed
  • Not disappointed
  • N/A, no longer use it

Sean Ellis proposed 40% "very disappointed" as a reference based on his work with startups. Treat it as a survey heuristic, not a validated boundary between PMF and no PMF. Report the respondent count and selection criteria, then pair the result with retention and customer research.4

2. Retention Curves

Plot cohort retention over time and interpret its shape alongside the product's expected usage cadence.

  • Flattening curve: Evidence that a group of users may be finding lasting value.
  • Curve to zero: A signal to investigate whether users are receiving ongoing value.

For a defined comparison, ChartMogul's 2025 billing report shows 2024 median annual customer retention of 41% for monthly-plan cohorts below $25 monthly ARPA and 69% for the $500–1,000 band. The chart excludes companies below $300,000 ARR and those with only one billing model. These are sample observations, not PMF thresholds or retention-curve flattening points.5

The appropriate timeframe depends on the product's usage cadence and customer lifecycle. Compare cohorts over a period that fits the product.

3. Growth Accounting and Quick Ratio

Quick Ratio = (New + Resurrected users) / Churned users.

Example: 2,000 new + 500 resurrected, 800 churned. Quick Ratio = 3.1.

Read Quick Ratio alongside retention, acquisition mix, and the definition of an active user. A value below 1 means churned users outnumber new and resurrected users in the period, but it does not diagnose the cause.

4. Organic Growth Percentage

Of last month's new users, how many came from non-paid sources? SEO, referrals, direct, communities.

Organic acquisition can be a useful signal, but the expected share depends on channel, category, and growth stage. Compare it with retention and customer research before drawing a PMF conclusion.

5. NPS, but Segmented

Ask "How likely are you to recommend us?" 0-10. Calculate % Promoters (9-10) minus % Detractors (0-6).

Segment NPS when groups have meaningfully different jobs, usage, or expectations. A blended score can conceal those differences.

A dating app might have NPS of 65 from singles and 10 from married users. A blended score of 35, depending on the sample mix, would obscure that difference. Investigate the relevant segment alongside retention and customer feedback before drawing a PMF conclusion.

Userpilot reports a median NPS of 39 and average of 35.7 for 229 B2B SaaS companies using its NPS Dashboard. The public page does not disclose the collection window or industry subgroup counts. Use it only as a qualified comparison, not a PMF cutoff.6

Qualitative Signals

These often appear before the metrics catch up.

Customer pull. Users verb your product ("Slack me," "DM me on X"). Creative workarounds to do more. Genuine anger during outages. Offers to pay before you ask.

Internal signals. Support shifts from "Why doesn't this work?" to "How can I do more?" Sales cycles shorten. Price stops being the main objection. Saying no to features becomes easy.

Market signals. Competitors copy. Investors reach out cold. Press covers without PR. Job posts list your product as a required skill.

Treat these as prompts to investigate with retention, customer research, and segment-level data.

Test Your PMF

Review the respondent count, sampling method, and results by meaningful segment. Read promoter and detractor feedback alongside observed behavior. NPS alone cannot establish PMF.

Review whether retention stabilizes and whether newer cohorts improve on older ones. An uptick may reflect returning users, depending on the retention definition. Check the calculation before interpreting the shape.

Comparing evidence across business models

Choose measures that reflect how customers receive value. For B2B SaaS, review logo retention and expansion; for mobile products, use retention intervals that fit the expected usage cadence. Marketplaces need evidence of transactions and repeat use, while enterprise products need evidence that pilots convert into continuing customer relationships.

The SaaS benchmark reference provides published comparisons with their populations and definitions. None establishes PMF on its own.

Superhuman: using survey segments to guide product work

In his account of Superhuman's process, Rahul Vohra reports that 22% of surveyed users initially said they would be very disappointed without the product in summer 2017. Focusing on a narrower set of customer personas raised that score to 33%. After three quarters of product improvements, it reached 58%.4

The team used feedback to identify what supporters valued and what held back users who were only somewhat disappointed. This informed a roadmap split between strengthening existing benefits and addressing missing capabilities. The sequence matters: segmentation changed the population being measured, while later product work addressed needs within that focus.

Common False Positives

These observations can be mistaken for stronger evidence than they provide.

Honeymoon phase. Early enthusiasm may not persist. Follow cohorts long enough to assess repeat value at the product's expected usage cadence.

Niche trap. Strong PMF inside a tiny, unscalable segment. Validate the TAM before betting the company.

One-feature PMF. Engagement is real for one feature, the rest of the product is a ghost town. Look at feature-level retention.

Paid growth mirage. New users grow with ad spend. Organic is flat. Track organic growth percentage and CAC payback. If paid retention is half of organic, investigate targeting, expectations, and the economics of that channel.

Enterprise pilot purgatory. Dozens of pilots, no conversions to multi-year contracts. Set strict conversion criteria and track them, otherwise pilots become a free consulting business.

AI Prompts for PMF Work

Use Claude, ChatGPT, or Gemini with the same grounding rule from the user research synthesis guide: cite the rows, do not paraphrase quotes.

Sean Ellis Survey Analysis

Analyze these PMF survey responses: [paste responses]

Calculate the % "very disappointed" score.
Segment by user characteristic if possible.
Identify the core persona of the "very disappointed" group.
Extract three themes from "somewhat disappointed."

Cite verbatim quotes for each theme. If you cannot find evidence, say so.

Retention Diagnosis

Cohort retention data: [paste]

Calculate and plot cohort retention from the supplied data. Identify whether each curve stabilizes.
Compare against [B2B / B2C] benchmarks.
Identify whether newer cohorts are improving.

PMF Segmentation

User data: [paste]

Cluster users by behavior or attribute. Calculate Sean Ellis or retention per segment. Rank by PMF strength. Profile the highest-PMF segment.

Suggest one targeting change.

Maintaining PMF

Market conditions, competitors, and the product can change, so continue checking whether customers receive lasting value.

Track these on a cadence:

  • Weekly: engagement metrics (DAU/MAU, session frequency)
  • Monthly: retention cohorts
  • Quarterly: Sean Ellis test on active users
  • Annually: market positioning and TAM review

Action Plan

Today (20 minutes). Send the Sean Ellis question to a sample of active users. A small sample can identify questions to investigate, but report the count and avoid treating it as a precise population estimate.

This week (2 hours). Build monthly cohort retention curves. Find the flattening point, or note that it doesn't flatten.

This month (1 day). Build a PMF dashboard with the four core metrics: Sean Ellis %, Day 30 retention, Quick Ratio, Organic growth %. Review it monthly.

This quarter. If you do not have PMF, design a four-week experiment with a clear new hypothesis. If you do have PMF, design the systems to maintain it.

Reassess PMF as customers, competitors, and the product change.

Sources

Footnotes

  1. Marc Andreessen, "The Only Thing That Matters" ↩

  2. CB Insights, "The Top 12 Reasons Startups Fail" ↩

  3. Sean Ellis, "Using Survey.io to Find Product/Market Fit" ↩

  4. Rahul Vohra, "How Superhuman Built an Engine to Find PMF," First Round Review ↩ ↩2

  5. ChartMogul, SaaS Billing Report 2025. Overall report covers 2,500 SaaS companies; subgroup counts for this chart are not published. ↩

  6. Userpilot, SaaS Product Metrics Benchmark Report. ↩